Table 3.

Performance of chosen two and three-point LLSs based on linear regression fitting and Bayesian estimation for prediction of exposure to treosulfan after different dosing regimens

Equation (if applicable)	R 2	PE > 20%	PE < 20%	MPE [%]	MAPE [%]	RMSE [%]
12 g/m2 in 1 h infusion
AUCpred = 0.86×C1 h + 1.93×C2 h + 7.47×C6 h – 42.2	0.9978	0	0	0.13	1.13	1.36
AUCpred = 2.78×C1.5 h + 6.62×C6 h – 45.6	0.9743	0	0	− 0.12	2.78	4.09
Bayesian estimation from C1 h, C2 h, C6 h	–	0	0	− 0.95	3.55	4.39
Bayesian estimation from C1.5 h, C6 h	–	0	0	− 0.36	6.58	7.86
12 g/m2 in 2 h infusion
AUCpred = 1.82×C2 h + 1.51×C3 h + 9.39×C8 h – 24.1	0.9993	0	0	0.11	0.78	0.93
AUCpred = 2.10×C2 h + 7.73×C6 h + 2.94	0.9929	0	0	− 0.29	1.40	1.67
Bayesian estimation from C2 h, C3 h, C8 h	–	0	0	− 0.61	4.35	5.60
Bayesian estimation from C2 h, C6 h	–	2	0	0.50	5.37	7.56
14 g/m2 in 2 h infusion
AUCpred = 2.01×C2 h + 2.15×C4 h + 7.68×C8 h – 9.07	0.9998	0	0	− 0.29	0.45	0.62
AUCpred = 2.07×C2 h + 7.57×C6 h + 42.6	0.9972	0	0	− 0.29	1.37	1.61
Bayesian estimation from C2 h, C4 h, C8 h	–	1	0	− 0.92	5.13	6.68
Bayesian estimation from C2 h, C6 h	–	1	0	0.24	5.97	8.12

LSS limited sampling strategy, R ² adjusted coefficient of determination, PE relative prediction error, MPE mean relative prediction error, MAPE mean absolute relative prediction error, RMSE root mean squared relative prediction error, AUC _pred predicted area under time-concentration curve, C _nh concentration of treosulfan measured n hours after the beginning of infusion